Physical activity contributes to the prevention of numerous chronic diseases . However, declining levels of physical activity have been reported in some developed countries despite considerable efforts to promote physical activity in different settings [2–4]. Environmental and policy interventions have been identified as the most promising strategies for achieving population-wide increases in physical activity . Cross-sectional and longitudinal studies examining associations between characteristics of the built environment and levels of cycling or walking  suggest that improving transport infrastructure in ways that favour active travel may help influence people to take up cycling or walking instead of using cars, thereby increasing their overall physical activity. However, little evidence has been gathered from intervention studies in which the effect of infrastructural improvements on physical activity has been measured [7, 8]; nor can such evidence easily be generated by researchers, who are rarely in a position to implement their own interventions in the built environment.
One way of addressing this lack of evidence is to conduct a natural experimental study to evaluate the effect of an intervention — in this case, a change to the environment involving improvements to transport infrastructure — that is not introduced for research purposes but is nonetheless amenable to evaluation . Where such events occur that give rise to variation in exposure to interventions, researchers should consider taking the opportunity to evaluate their effects using robust, practical and cost-effective measures . Guidance from the National Institute for Health and Clinical Excellence (NICE)  recommends that intervention studies of this kind should include valid measures of physical activity before and after the intervention to test associations between changes to the physical environment and changes in physical activity. In a natural experimental context, having no control over the implementation of an intervention sometimes constrains researchers to a limited time period for baseline data collection. In such circumstances, it is often easier to rely on the most commonly-used approach to measuring physical activity, which is to use self-reported measures . However, it may sometimes be possible to incorporate objective measurement of baseline physical activity using devices such as accelerometers, even if the time available for data collection is limited.
A search of PubMed for studies using accelerometers to measure physical activity and published between 2005 and 2010 retrieved around 100 studies. More than half were cross-sectional studies, with sample sizes ranging from less than 100 to more than 2000. An example of a study at the upper end of this range is the Health Survey for England in 2008, in which 2115 adults were reported to have returned accelerometer data of a satisfactory standard for analysis . Longitudinal studies have mostly been conducted in children, with studies such as ALSPAC  and SPEEDY  having collected accelerometer data from more than 1000 participants; however, certain longitudinal studies in adults such as the Nakanojo  and NHANES  studies have collected accelerometer data from more than 3000 participants, albeit not necessarily with more than one wave of accelerometer measurement. Relatively few intervention studies have been reported. The largest intervention study found in this search was that of a school-based intervention to reduce the prevalence of overweight in a sample of more than 3135 children and adolescents . In this study, a total of 1538 participants from both intervention and control schools were randomly selected for objective physical activity measurement in weekly batches over a four-year intervention period. The largest intervention study among adults was a clinic-based behavioural intervention involving 236 women , of whom 178 were measured at baseline and followed up after six months and 173 were measured again after 12 months. The search found little evidence that objective physical activity measurement had been used in natural experimental studies, in which researchers have no control over the intervention. There is also little evidence-based guidance on how best to deploy accelerometers for the assessment of free-living physical activity in large studies .
A few studies have attempted to evaluate the effect of environmental interventions on active travel, but many have not included overall physical activity as an outcome  as recommended by NICE  and of those that have, few have incorporated objective measures of overall physical activity. For example, the RESIDE study has used survey and pedometer data to evaluate the impact of the Department of Planning’s Liveable Neighbourhood guidelines on the health and active travel of people moving into new homes in Western Australia , and in the UK the M74 [21, 22] and iConnect  studies have used or adapted the short version of the International Physical Activity Questionnaire (IPAQ)  for baseline measurement, with iConnect including accelerometry only for specialist case studies. Studies of this kind sometimes encounter unexpected circumstances during the implementation of the intervention which require a high degree of flexibility on the part of researchers and, sometimes, of funding bodies. The challenges of completing baseline accelerometer measurement on a large scale in a limited time while maintaining a high level of data quality are therefore likely to be encountered by other researchers conducting similar studies in the future. In this study, we attempted the rapid collection of baseline accelerometer data from a large number of participants by post, without face-to-face contact. The aim of this paper is to report and reflect on the practical issues, challenges and results of this exercise in rapid baseline objective physical activity measurement in a natural experimental study.